Color contrast describes the influence of one color on the perception of colors in neighboring areas. This study addressed two issues: (1) the accurate representation of the color changes; (ii) the underlying visual mechanisms. Observers matched the hue that was induced in a neutral square when it was set in one of four standard colored surrounds: “red” (+L(−M) relative to neutral), “green” (−L(+M)), “purple” (+S), and “yellow” (−S). The standard and matching displays were viewed haploscopically. The standard neutral square was either a luminance increment, or decrement, both of which appeared the complementary color to the surrounds in which they were inset. In Experiment 1, the surround luminance in each eye's display was either equal, at 18 cd·m−2, or the match surround luminance was reduced to 2.5 cd·m−2. The matches with equal surround luminances could be represented as vector shifts in a logarithmic MacLeod–Boynton (r, b) chromaticity diagram, as described previously (Shepherd, 1997, 1999). The low luminance matches were, however, displaced further from neutral, as if larger chromatic differences were needed. The precise direction of the displacements differed for luminance increments and decrements: the red, green and yellow decrement matches were also displaced vertically downwards in the MacLeod-Boynton diagram. In Experiment 2, dark-adapting before setting repeat color matches displaced the decrement matches vertically, but did not affect the increment matches. Thus, rod intrusion in S-cone pathways may have boosted the S-cone signal for the lowest luminance decrement matches in Experiment 1 and account for the vertical shift in MacLeod-Boynton co-ordinates. The distinct pattern of displacements for low luminance increments and decrements may be explained if the match is set at a cone-opponent, rather than a cone contrast, site and if rod signals have an input only to S-cone decrement, perhaps S-OFF, pathways.